Molecular Design of Unimolecular Tetra-receptor Agonists

Abstract

Peptide hormone-receptor interactions serve as critical regulators of metabolic homeostasis, a paradigm exemplified by the clinical efficacy of glucagon-like peptide-1 (GLP-1) receptor agonists. Building upon this framework, strategic design has yielded unimolecular dual and triple agonists targeting GLP-1R, glucose-dependent insulinotropic polypeptide receptor (GIPR), and glucagon receptor (GcgR), leveraging the sequence homology within the cognate native ligands of the class B G protein-coupled receptor (GPCR) family. However, the integration of Y2 receptor (Y2R) agonism—engaged by peptide YY (PYY) and belonging to the structurally divergent class A GPCR family—has remained an unaddressed challenge due to the topological and sequence disparities between these receptor classes. Y2R activation plays a pivotal role in appetite suppression, potentiating the metabolic benefits conferred by GLP-1R, GIPR, and GcgR agonism. Here, we report first-in-class, unprecedented tetra-agonists with high potency at GLP-1R, GIPR, GcgR, and Y2R. The chimeric peptides overcome the intrinsic sequence constraints imposed by class A and class B GPCR divergence, demonstrating the feasibility of rationally designed agonism mediated by single agents across receptor families. Lipidation of this template is well tolerated enhancing the promise of therapeutic viability. Furthermore, we show that biased agonism at GLP-1R selectively boosts cyclic AMP (cAMP) signaling while minimizing β-arrestin recruitment, thereby decoupling receptor desensitization from metabolic efficacy. Additionally, we introduce a tunable framework to modulate β-arrestin engagement without compromising cAMP potency, providing insight into the fine-tuning of GPCR-mediated signaling for next-generation peptide therapeutics.

Graphical Abstract

INTRODUCTION

The therapeutic utility of glucagon-like peptide-1 receptor (GLP-1R) agonists in glycemic control has been well established. Activation of GLP-1R potentiates glucose-dependent insulin secretion from pancreatic β-cells while concurrently inducing increased satiety and delayed gastric emptying (Fig. 1a).¹ To enhance the pharmacokinetic profile of GLP-1-based therapeutics, various chemical modifications have been employed to maintain receptor potency while preventing hydrolysis catalyzed by the ubiquitous serine protease, dipeptidyl peptidase-4 (DPP4).^2–7 Semaglutide, a widely used GLP-1R mono-agonist, incorporates 2-aminoisobutyric acid (Aib, X) at the second position from the N-terminus to confer proteolytic resistance while employing a lipid diacid side chain to extend circulatory half-life via serum albumin binding, ascribed to diminished glomerular filtration.⁸

Figure 1.

(a) Receptor mediated actions initiated by peptide hormones highlighted in various organs/tissues. Endogenous native agonists of the GLP-1R (blue, GLP-1), GIPR (red, GIP), GcgR (purple, glucagon, Gcg), and Y2R (green, PYY) and their relevant sites of action. Stimulatory and inhibitory effects are indicated by upwards and downwards arrows, respectively. (b) Unimolecular multi-receptor agonists are chimeric peptides that blend the sequences of GLP-1, GIP, and glucagon based in part on their high sequence homology particularly in the N-termini of these peptides which confer cognate receptor agonism. (c) PYY acts as an agonist at the Y2 receptor (Y2R), with its C-terminal region playing a critical role in receptor activation. White colored amino acid residues represent regions of the respective constructs that were cut out to engineer hormone tetra-receptor agonist chimeras (TCs). (d) A first of its kind unimolecular TC, including N-terminal modifications to protect against degradation by DPP4 as well as peptide lipidation with the potential to delay drug clearance from the circulation.

Expanding beyond GLP-1R mono-agonists, the role of glucose-dependent insulinotropic polypeptide (GIP) and glucagon (Gcg) in metabolic regulation has garnered burgeoning interest. GIP exerts regulatory effects on lipid metabolism and adiposity while acting on pancreatic β-cells via GIP receptor (GIPR) agonism, and Gcg receptor (GcgR) activation enhances energy expenditure (Fig. 1a).⁹ Several studies have demonstrated that co-administration of GLP-1R, GIPR, and GcgR agonists synergistically improves glycemic control and metabolic outcomes, reinforcing the rationale for developing unimolecular multi-receptor agonists.⁹

Given the conserved structural and functional homology among GLP-1R, GIPR, and GcgR—all of which belong to the class B family of G protein-coupled receptors (GPCRs)—peptide chimeras capable of simultaneously stimulating these receptors have emerged as a compelling pharmacological strategy (Fig. 1b, Fig. 2a). The initial enthusiasm surrounding these unimolecular chimeric co-agonists stemmed from the development of GLP-1R/GcgR and GLP-1R/GIPR dual agonists, which demonstrated stable glycemic control and enhanced weight reduction compared to GLP-1R mono-agonists.^{10, 11} Tirzepatide, a clinically approved GLP-1R/GIPR dual agonist, elicits a 15–20% reduction in body weight, consistently outperforming agonists that only stimulate GLP-1R.^12–14 This enhanced weight loss may be partially attributed to GIPR agonism, which has been shown to mitigate the nausea commonly associated with GLP-1R activation.^{15, 16} By reducing nausea, dual agonism at GLP-1R and GIPR may enable higher tolerated doses, thereby amplifying weight loss outcomes. The therapeutic potential of further expanding the range of targeted receptors was demonstrated by Finan et al., who reported that combining a GLP-1R/GIPR dual agonist with a glucagon analogue yielded superior weight loss in diet-induced obese mice compared to dual agonism alone.⁹ This finding catalyzed the rational design of triagonists capable of activating GLP-1R, GIPR, and GcgR within a single molecular entity.^{9, 10, 17} Notably, LY3437943 (retatrutide),¹⁸ a triagonist of GLP-1R, GIPR, and GcgR, recently demonstrated a maximal 24% weight loss in overweight and obese individuals in a 24-week Phase 2 clinical trial.¹⁹ There appears to be a synergistically driven increase with maximal weight loss driven by the number of receptors agonized, reinforcing the rationale for developing multi-receptor agonists that extend beyond current triagonist paradigms.

Figure 2.

(a) Peptide sequence of native cognate agonists of Y2R (PYY), GLP-1R (GLP-1), GIPR (GIP), and GcgR (glucagon, Gcg). Previously reported synthetic analogues of one or multiple receptor agonists are shown with sequence homology (yellow highlight), noncanonical amino acids (blue and green), and lipid side chains (red box). The ‘exendin-tail’ sequence included in dual (Tirzepatide) and multi-agonists (triagonist and retatrutide) is underlined. (b) Cryo-EM structure of PYY_3–36 bound to the Y2R (grey ribbon): Only the PYY_24–36 C-terminal (‘ct’); 13 amino acids are shown (structure and sequence in green) with residue Leu24 exiting the binding groove (PDB: 7YON). (c) Tetra-chimeric (TC) multi-agonists that are (i) derived from residues 1–32 of triagonist or retatrutide and (ii) contain the PYY_24–36 (ct) sequence, with the “GGPS” exendin-tail segment acting as a spacer between components (i) and (ii). (d) Various linker-lipid motifs on the lysine side chain utilized combination(s) of β-alanine (βAla, βA), ɣ-L-glutamic acid (ɣGlu, ɣE), and 8-amino-3,6-dioxaoctanoic acid (OEG, O), and lipid diacids or C16 acyl lipids. (e) Lipid side chains used in a and c. (f) Noncanonical amino acids present in literature compounds or those described in this work.

Peptide YY (PYY), a 36-residue hormone, has recently emerged as another key regulator of metabolism with significant therapeutic potential in weight management.^20–23 PYY circulates in two biologically active forms: the full-length peptide (PYY1–36), which engages multiple neuropeptide Y (NPY) receptors, and the truncated DPP4-cleaved form (PYY3–36), which selectively agonizes the neuropeptide Y2 receptor (Y2R) (Fig. 1a, c).²⁴ Y2R activation has been shown to reduce food intake and promote weight loss via central satiety signaling,^{25, 26} delayed gastric emptying, and enhanced lipolysis and thermogenesis.^{24, 25} Co-administration studies have demonstrated that combined Y2R and GLP-1R agonism produces greater improvements in HbA1c, weight loss, and insulin secretion compared to either agonist alone.²⁷ Unlike GLP-1R, GIPR, and GcgR, which critically depend of the N-termini of cognate endogenous peptide agonists for receptor activation, Y2R—belonging to the class A GPCR family—interacts predominantly with the C-terminal region of the ligand (Fig. 2b).²⁴ Structural analyses of the PYY3–36:Y2R complex have revealed that only a small C-terminal region of PYY engages with the putative Y2R agonist binding pocket.²⁸ The potential of Y2R agonism has been further supported by Østergaard and coworkers, who developed a dual GLP-1R/Y2R agonist, which demonstrated enhanced food intake suppression.²⁹ Moreover, a combination therapy study incorporating a long-acting GIPR agonist and a Y2R agonist confirmed that Y2R activation induces weight loss while additional GIPR agonism attenuates nausea, a common side effect of Y2R agonists.³⁰

Obesity and type 2 diabetes (T2D) continue to pose major global health challenges, with their increasing prevalence driving significant morbidity and mortality. The close interconnection between these metabolic disorders, collectively referred to as ‘diabesity,’ underscores the urgent need for novel therapeutic strategies that simultaneously address weight management and glycemic control.³¹ Given the promising metabolic benefits conferred by GLP-1R, GIPR, GcgR, and Y2R agonism, we hypothesized that the C-terminal domain of PYY could be strategically grafted onto carefully optimized chimeric triagonists to generate a novel class of unimolecular tetra-receptor agonists.

Here, we report the development of a series of tetra-hormone chimeras (TCs) that integrate the C-terminal Y2R agonist sequence with N-terminal peptide determinants of GLP-1R, GIPR, and GcgR activation. These TC constructs were rationally designed toward retaining high potency and efficacy across all four target receptors while remaining within a sequence length that is amenable to solid-phase peptide synthesis (SPPS) and scalable manufacturing methods (Fig. 1b–d). The ability to achieve broad-spectrum receptor activation within a single peptide framework represents a promising advancement in the field of metabolic therapeutics, with implications for the treatment of obesity, T2D, and associated comorbidities.

RESULTS

Rational Design of tetra-agonists

The peptide hormones GLP-1, GIP, glucagon, and PYY exhibit high affinity and specificity for their respective receptors (GLP-1R, GIPR, GcgR, and Y2R) (Ref. 9), with little if any cross-reactivity (Fig. S1). To explore the feasibility of designing a single peptide capable of significantly engaging and activating all four receptors, we began by examining a previously developed triagonist (peptide 20 in their study) by Finan, DiMarchi, and colleagues.⁹ This peptide was engineered to potently co-activate the GLP-1R, GIPR, and GcgR (“Triagonist”, Fig. 2a). This agent achieves a balanced activation of all three target receptors through a series of carefully designed structural modifications. To enhance its receptor affinity and functionality, the triagonist underwent extensive side chain modifications aimed at integrating key sequence traits from GLP-1, GIP, and Gcg. These optimizations facilitated activation of all three receptors by enabling a more seamless and independent interaction with each. This work also established the necessity of incorporating the C-terminal 11 amino acid sequence of exendin-4 (“exendin-tail”), a feature now prevalent in many subsequently reported multi-agonists. Binding studies with the above triagonist confirmed that this compound has negligible affinity for Y2R (and for several other unrelated receptors), consistent with the known requirement of the PYY C-terminal sequence for binding and activating the latter receptor.⁹ To impart Y2R activity to the triagonist while maintaining its agonism at GLP-1R, GIPR, and GcgR, we strategically modified the lipid acylation site of the template molecule, replacing it with a tyrosine residue at position 10. Additionally, we engineered the peptide’s C-terminal region by substituting residues 33–39 with the corresponding segment from peptide YY (PYY), specifically incorporating residues 24–36 of PYY. This design refinement preserves receptor selectivity while introducing new pharmacological properties that enhance its functional repertoire. This judiciously selected truncation of the N-terminal domain (residues 1–32) of the TC template was informed by examination of cryo-EM structures of class B GPCR agonists bound to the GLP-1 and GIP receptors. In all agonist–receptor complexes, residues 30–33 of the agonist were solvent-exposed and did not contact the receptor (GLP-1:GLP-1R, PDB: 6X18; GIP:GIPR, PDB: 7RA3; semaglutide:GLP-1R, PDB: 7KI0; tirzepatide:GLP-1R, PDB: 7RGP; tirzepatide:GIPR, PDB: 7RBT). We hypothesized that retaining a minimal exendin-tail sequence of the template triagonist, i.e. the “GGPS” motif, would serve as a structural spacer, separating the N-terminal segment responsible for GLP-1R, GIPR, and GcgR activation from the newly introduced Y2R-targeting domain. Including such a spacer could preserve affinity for the receptors targeted by the original Finan/DiMarchi triagonist.

Previous studies have demonstrated the critical role of the PYY C-terminal region in Y2R activation (Østergaard et al., Ref. 29). Additionally, research by Roth et al. demonstrated that appending the PYY C-terminus at position 33 in a truncated exendin-4 (1–32) derivative was structurally and functionally viable.³² These insights guided the development of our modified derivative, TC0 (Fig. 2c), which demonstrated Y2R agonism, albeit with a 6.7-fold decrease in potency compared to native PYY. TC0 retained partial GcgR activity (10-fold reduction compared to Gcg) and exhibited strong GLP-1R activation, with a 6.5-fold potency increase over native GLP-1. However, TC0 exhibited a significant loss in potency at the GIPR (80-fold lower than GIP), likely due to truncation of the exendin-tail truncation, which Finan et al. previously identified as essential for GIPR engagement. Overall, TC0 may be considered as a functional triagonist for GLP-1R, GcgR, and Y2R as it exhibits suboptimal GIPR engagement. The latter limitation highlights the need for further optimization of TC0, or for exploring alternative templates such as retatrutide to achieve true tetra-agonist functionality (Table 1).

Table 1.

Potencies derived from G-protein mediated cAMP-dependent signaling of various agonists at the four target receptors.

		GLP-1R		GIPR		GcgR		Y2R
Peptidea	Modificationb	EC50 (pM)c	Fold-Shift (↑/↓)×e	EC50 (pM)c	Fold-Shift (↑/↓)×e	EC50 (pM)c	Fold-Shift (↑/↓)×e	EC50 (pM)c	Fold-Shift (↑/↓)×e
Cognate native agonistf	-	2.6	-	3.5	-	12	-	60	-
Semaglutide	-	2.6	equal	-	-	-	-	-	-
Tirzepatide	-	35	↓ 13	2.6	equal	-	-	-	-
Retatrutide	-	11	↓ 4.2	2.2	↑ 1.6	8.2	equal	-	-
GEP44	-	0.64	↑ 4.1	-	-	-	-	>104 d	↓ 200
TC0	Triagonist(Y10, 33-PYYct)	0.4	↑ 6.5	280	↓ 80	120	↓ 10	400	↓ 6.7
The following are derived from the retatrutide scaffold
RET1	Retatrutide (L13, Q20)	36000	↓ 14,000	16000	↓ 4,600	3600	↓ 300	-	-
RET2	RET1 (K17)	3.1	equal	1200	↓ 340	3500	↓ 290	-	-
TC1	RET2 (33-PYYct)	0.4	↑ 6.5	2.3	↑ 1.5	130	↓ 11	170d	↓ 2.8
TC2	TC1 (K10[3], V37, W39, L40)	2.1	equal	0.47	↑ 7.4	30	↓ 2.5	55	equal
TC2a	TC2 (K10[ɣE-C20DA])	17	↓ 6.5	120	↓ 34	3200	↓ 270	1200d	↓ 20
TC2b	TC2a (Y10, K31[2])	110	↓ 42	200d	↓ 57	ND		160d	↓ 2.7
TC2c	TC2b (K31[4])	95	↓ 37	-	-	-	-	330	↓ 5.5
TC2d	TC2c (K16[2], P31)	89	↓ 34	39	↓ 11	440	↓ 37	440	↓ 7.3
TC2e	TC2d (K16, K17[1])	91	↓ 35	2.7	equal	160	↓ 13	170	↓ 2.8
TC2f	TC2e (K17[2])	89	↓ 34	5.6	↓ 1.6	250	↓ 21	290	↓ 4.8
TC2g	TC2f (K17[4])	55	↓ 21	6.3	↓ 1.8	140	↓ 12	1700d	↓ 28
TC2h	TC2g (K17[1], X20)	47	↓ 18	-	-	-	-	-	-
TC2i	TC2h (K17[O2X-βA-ɣE-C18DA])	80	↓ 31	74	↓ 21	-	-	3100d	↓ 52
TC2j	TC2i (K17[O-βA-ɣE-C18DA])	47	↓ 18	1.3	↑ 2.7	21d	↓ 1.8	1900	↓ 32
TC3	TC2j (K17[5])	1.1	↑ 2.4	0.75	↑ 4.7	23	↓ 1.9	420	↓ 7
TC3a	TC3 (K17, NMeR44)	4.7	↓ 1.8	-	-	-	-	1800d	↓ 30
TC3b	TC3a (K17[1])	57	↓ 22	-	-	-	-	-	-
TC3c	TC3b (K17[5])	20	↓ 7.7	1.1	↑ 3.2	29	↓ 2.4	>104	↓ 270
TC3d	TC3c (Lα13, K17)	1.1	↑ 2.4	0.47d	↑ 7.4	-	-	-	-
TC3e	TC3d (K17[1])	21	↓ 8	0.95	↑ 3.7	38d	↓ 3.2	-	-
TC3f	TC3e (K17[5])	18	↓ 7	1.2	↑ 2.9	41	↓ 3.4	>104	↓ 330
TC4	TC3 (R16, A20)	31	↓ 12	0.98	↑ 3.6	81	↓ 6.8	1100	↓ 18

Potencies at the GLP-1R, GIPR, and GcgR were determined using cAMP-dependent CRE_6x-luciferase reporter gene assays (see SI). Potencies at the Y2R (inhibition of cAMP production) were measured using the chimeric G_qi5 induced SRE-luciferase reporter system. All assays were performed in HEK293 cells.

^aCompounds previously reported and new constructs derived from either a known triagonist (middle of dotted lines, TC0) or from retatrutide as a scaffold (below dotted lines). Key peptide sequences are shown in Figure 2.

^bCompound modifications (shown in parentheses) from prior templates as indicated. Lipid side chain modifications (in brackets) are acylated on lysine residues at the indicated position.

^cEC₅₀ is the concentration of peptide required for half-maximal activity at the target receptor and is shown as (–) when not tested. Experiments where no agonism was detected are labeled (ND). Unless otherwise indicated, a minimum of three separate experiments were performed for each peptide at each respective receptor. Variability of EC₅₀ determinations, represented by the standard error of the mean (SEM) is reported in Table S1.

^dMean EC₅₀ of a compound where two separate experiments were performed.

^eFold-shift compares the potency of each compound to that of the native ligand at its cognate receptor [more (↑) or less potent (↓)]. A difference of ≤ 1.5-fold is considered equally active at each receptor.

^fThe cognate native ligands used were GLP-1 for GLP-1R, GIP for GIPR, Gcg for GcgR, and PYY for Y2R.

To pursue the latter option, and to systematically explore the modular adaptability of our unimolecular multi-receptor agonist platform, we then focused on LY3437943 (retatrutide) from Eli Lilly as a template for chemical modification. Functioning as a triagonist, LY3437943 stimulates GLP-1R, GIPR, and GcgR with balanced potency and incorporates a diacid lipid moiety to enable once-weekly administration.^{18, 19} This construct also features strategically placed noncanonical amino acids to enhance metabolic stability and optimize receptor interaction. Specifically, 2-aminoisobutyric acid (Aib) at position 2 confers resistance towards DPP4 catalyzed hydrolysis and inactivation, while α-methyl-L-leucine (αMeLeu, Lα) at position 13 and an additional Aib at position 20 contribute to enhanced receptor agonism and enzymatic stability (Fig. 2a).

While these design elements offer pharmacological benefits, incorporating αMeLeu at position 13 and Aib at position 20 complicates manufacturing due to the need for chemical synthesis and the high cost and low coupling efficiency of substituting these residues.³³ To mitigate these limitations while preserving the functional integrity of the parent scaffold, we sought to evaluate substitutions that eliminate dependence on these noncanonical residues. We hypothesized that a dual αMeLeu13L and X20Q change would retain receptor engagement while streamlining manufacturing efforts. The resulting construct, RET1, was designed to assess the extent to which these modifications influence receptor stimulation and signaling.

Unexpectedly, RET1 exhibited a pronounced loss of agonistic activity at GLP-1R, GIPR, and GcgR, despite bearing the diacid lipid moiety and preserving the overall structural integrity of the parent scaffold (Table 1). This unexpected result suggested that the two substituted noncanonical amino acids present in retatrutide—previously assumed to be auxiliary stabilizing elements—were, in fact, critical determinants of receptor engagement. Their absence severely diminished activity across all three target receptors, casting doubt on the feasibility of this simplified scaffold as a viable triagonist.

Lipidation plays a key role in peptide pharmacokinetics and receptor binding by modulating conformation and bioavailability. To evaluate the contribution of this peptide modification to receptor engagement, we generated RET2 by excising the diacid lipid moiety at position 17. Removal of this modification resulted in enhanced GLP-1R activation compared to RET1, suggesting that lipidation may influence receptor binding dynamics in not fully understood ways. However, RET2 failed to restore GIPR and GcgR activation, demonstrating that lipid excision alone was insufficient to re-establish multi-receptor agonism (Table 1).

Given the incomplete recovery of RET2, we sought an alternative strategy to compensate for the elimination of noncanonical residues. We truncated RET2 at amino acid 32 and substituted the 13-amino acid PYY C-terminal domain at position 33, mirroring the modification employed in TC0, to determine whether this addition could simultaneously confer Y2R agonism and restore GIPR and GcgR activity. This rational design approach aimed to explore the modular adaptability and chimeric nature of the scaffold, while testing whether strategic grafting of a Y2R-targeting motif could reinstate functional engagement across all four receptors.

To our pleasant surprise, the resulting construct, TC1, exhibited a more desirable functional profile distinct from its predecessors. TC1 successfully re-established agonism at GIPR and GcgR while simultaneously conferring Y2R activation, akin to TC0, but without reliance on noncanonical amino acids.³⁴ Compared to native endogenous agonists of cognate target receptors, TC1 demonstrated a 6.5-fold increase in GLP-1R potency, near-equivalent activity at GIPR, an 11-fold reduction in GcgR activation, and a 2.8-fold decrease in Y2R potency (Fig. 2a, Table 1). These findings challenge the prevailing assumption that the exendin-tail—a conserved structural feature in most GLP-1R/GIPR/GcgR triagonists reported to date—is an obligatory component for optimal receptor engagement.^{9, 17, 18, 35, 36} Instead, the integration of the PYY24–36 C-terminal domain appears to be a viable alternative for conferring favorable activity across the above receptor targets while at the same time adding robust Y2R agonism.

To further contextualize the receptor activity of our constructs, we synthesized and evaluated GEP44, a previously reported multi-receptor agonist that was of interest due to incorporated PYY sequence.^{32, 37} In our hands, the measured potency of GEP44 at Y2R was minimal, in contrast to published data (Table 1). Consistent with our findings, Østergaard et al. previously characterized a structurally related analogue (to GEP44) and reported weak Y2R binding, therefore proposing that the compound primarily functions as a GLP-1R mono-agonist.²⁹ Still, it is possible that discrepant findings on the function of GEP44 may, at least in part, reflect the use of different methodologies for measuring cAMP-inhibitory signaling via Y2R in our vs. other laboratories (via a recombinant inhibitory Gα5i construct vs. via endogenous Gαi).

To address this consideration, we conducted additional control experiments to evaluate the equivalency of detecting Y2R function via recombinant vs. endogenous G proteins. The chimeric recombinant G-protein assay we routinely used for quantifying agonist-induced Y2R function has been previously established as a robust platform for characterizing Gαi-coupled receptors in reporter gene assays.³⁸ To further validate this approach, we confirmed Y2R engagement through parallel assessment of endogenous Gαi-mediated inhibition of cAMP-dependent 6xCRE luciferase activity (Fig. S2). Both detection strategies (via recombinant Gq5i or via endogenous Gαi) yielded comparable results, with slightly higher detection sensitivity of the former strategy. The strong correlation between these assays, and consistency of our findings with established Y2R pharmacology of known receptor subtype-selective PYY variants, underscores the reliability of our experimental approach for assessing receptor activation by newly discovered TC compounds.

Collectively, our initial findings with retatrutide-derived chimeric peptides suggest a promising avenue for rational scaffold engineering towards generating optimized multi-receptor agonists. TC1 represents a pivotal advancement in the development of tetra-agonist peptides, setting a precedent of successfully integrating Y2R activity while preserving robust tri-receptor engagement, all with an option to avoid the synthetic constraints imposed by noncanonical amino acids. On this basis, we began iterative TC optimization to explore the range of feasible fine-tuning receptor selectivity, biased signaling, and initial consideration of compound stability.

Choice and positioning of lipids

Lipidation has emerged as a powerful strategy for enhancing peptide pharmacokinetics, particularly by leveraging serum albumin binding to extend circulation time, and to protect against enzymatic degradation (together leading to peptide lifetime “protraction”). Furthermore, peptide lipidations can be differentially tolerated at target receptors of multi-agonists and thereby modulate the potency preference of such compounds for individual receptors. Accordingly, we explored lipidation within the TC1 scaffold to modulate receptor engagement and stability. The initial modification involved conjugation of ɣ-L-glutamic acid (ɣGlu, ɣE) to a lysine residue, followed by attachment of a palmitic acid (C16 acyl) moiety, yielding lipid 3 (Fig. 2d, e). This motif, present in the triagonist peptide and liraglutide, has demonstrated affinity for albumin, with the potential to enhance pharmacokinetic properties.^{8, 39, 40} Careful examination of cryo-EM structures of the triagonist peptide (Peptide 20, P20) bound to GLP-1R, GIPR, and GcgR revealed that lipid 3 (Fig. 2), positioned at residue 10, is oriented toward the plasma membrane (P20:GLP-1R, PDB 7VBH; P20:GIPR, PDB 7FIN; P20:GcgR, PDB 7V35).⁴¹ We hypothesized that this membrane-facing orientation could serve as an anchor, potentially enhancing the peptide’s activity.⁴¹ In parallel, we introduced C-terminal modifications (L37V, L39W, and V40L) guided in part by literature precedent to enhance Y2R potency.²⁹ The resulting construct, TC2, featuring lipidation at position 10 and a modified PYY C-terminus, exhibited equipotent activity at GLP-1R, a 3.6-fold increase at GIPR, a 2.5-fold decrease at GcgR, and comparable Y2R potency (all relative to native corresponding native ligands at cognate receptors, Table 1 and Fig. 3a). Notably, TC2 displayed enhanced activity at both GcgR and Y2R compared to TC1, likely attributable to lipidation (for Gcg) and C-terminal modifications (for PYY).

Figure 3.

(a) Chemical structures of the tetra-peptide chimeras TC2 and TC3. Chimera TC4 is identical to TC3, except for two substitutions (Lys16Arg and Aib20Ala) made on the TC3 scaffold. (b) A representative experiment illustrating concentration-response curves from 6xCRE-luciferase reporter gene assays detecting cAMP production stimulated via GLP-1R, GIPR, or GcgR, and from Gq_i5 / SRE-luciferase reporter gene assays detecting inhibition of cAMP production via Y2R. (c) Cartoon depiction of tetra-peptide chimeras with arrows illustrating -fold agonist potency increase (↑) or loss (↓) compared to cognate native ligand at each targeted receptor. TCs are represented by symbols as in panel (a), and -fold potency changes are based on ≥ n = 3 independent experiments. (d) Bar graph potency comparison of cognate native receptor agonists (defined as 100%), the FDA-approved semaglutide, tirzepatide, and retatrutide, and the TC2, TC3, and TC4 chimeras. Relative % potency at each receptor = (native agonist EC₅₀/compound EC₅₀) × 100 from the data presented in Table 1 or previously published.⁹ Low/marginal potencies with EC₅₀ ≥ 100 nM are shown as ~0. Data in (b) and (d) represent the mean ± SEM.

These findings prompted further optimization of the lipid/protractor moiety, leading us to explore lipid diacids, which confer peptides with greater albumin affinity and prolonged in vivo circulation compared to monoacid lipids.⁸ Prior studies by Lau et al. demonstrated that semaglutide exhibited a >3-fold longer half-life than liraglutide in mini-pigs, which correlated with a 5.6-fold tighter albumin binding, as judged by analytical ultracentrifugation.⁸ Clinical formulations reflect this difference, with semaglutide-based drugs (Ozempic, Wegovy) requiring weekly administration, while liraglutide-based counterparts (Victoza, Saxenda) necessitate daily dosing. As a caveat, the longer half-life of semaglutide may not be entirely attributable to lipid protraction, as this compound is also protected from rapid degradation by N-terminal sequence modification (which is absent in liraglutide). To evaluate lipid diacid (DA) compatibility within the TC2 scaffold, we systematically scanned several positions. Replacing the C16 acyl at position 10 with C20DA (TC2a: K10[ɣE-C20DA]) resulted in substantial activity loss at the GIPR and GcgR, possibly due to unfavorable electrostatic interactions between the terminal acid and the cell membrane in the receptor-bound state. To mitigate this, position 10 was substituted with tyrosine to increase sequence homology with GIP and Gcg, while positions 31 and 16 were explored as alternative sites for lipidation. Lipidation at position 31, inspired by tirzepatide (TC2b) and retatrutide (TC2c), was tolerated at Y2R but resulted in >40-fold reductions in GLP-1R and GIPR activity, with no detectable GcgR activity (Table 1). However, lipidation at position 16 (TC2d), a solvent-exposed region in cryo-EM structures, and a known lipidation site in next-generation triagonists,³⁵ yielded improved receptor activity but did not fully recover to the potencies observed with the earlier TC2 template (Table 1). Given these findings, we investigated lipidation at position 17, a site utilized in retatrutide and in liraglutide derivatives.^{18, 42} This modification was tolerated at GIPR and GcgR, but GLP-1R activity remained below desirable levels (TC2(e-g), Table 1).

To address this, we introduced an Aib residue at position 20 to reinforce α-helical structure, a strategy frequently employed in multi-agonist design to stabilize bioactive conformations.^{13, 14, 33} We then examined lipid diacid tolerance in this modified construct, varying linker compositions including ɣGlu, 8-amino-3,6-dioxaoctanoic acid (OEG, O), and β-alanine (βAla, βA) at position 17. Among these, the βA-O-ɣE-C18DA motif (lipid 5) conferred high potency across all target receptors, yielding the TC3 scaffold. TC3 displayed modestly improved GLP-1R potency, a 4.7-fold enhancement at GIPR, a 1.9-fold reduction at GcgR, and a 7-fold reduction at Y2R (all relative to corresponding native ligands at cognate receptors, Table 1 and Fig. 2a). Notably, reversing the placement of β-alanine and OEG (TC2j) led to inferior GLP-1R and Y2R activity, underscoring the significance of linker orientation.

Encouraged by these findings, we explored additional enzymatic stabilization strategies. The N-methyl arginine (NMeR) modification at position 44, previously reported to enhance PYY stability,⁴³ was incorporated but resulted in Y2R activity loss. Non-lipidated TC3a exhibited a 30-fold reduction at Y2R, while lipid 5-bearing TC3c suffered a more dramatic 270-fold reduction relative to PYY. In parallel, we examined the αMeL modification at position 13, previously employed in retatrutide, to protect against neprilysin degradation.⁴⁴ While TC3a and TC3d retained high GLP-1R potency, lipidation to mimic the one applied to semaglutide (TC3b, TC3e) slightly diminished activity. Replacing the semaglutide lipid with lipid 5 (TC3c, TC3f) restored GLP-1R potency in TC3c, but did not impact TC3f, suggesting a context-dependent effect of αMeL incorporation.

To facilitate a future option for scalable production, we further refined the scaffold. The Aib residue at position 20 was substituted with alanine, and lysine at position 16 was replaced with arginine (X20A, K16R), yielding TC4 (Fig. 2a, Table 1). This design aligns with semaglutide’s yeast-based semi biosynthesis,⁴⁵ wherein a single lysine permits orthogonal lipidation, followed by N-terminal dipeptide coupling. Studies have also demonstrated feasibility of post-synthesis C-terminal amidation.⁴⁶ The resultant TC4 construct retained reasonable to excellent activity at all receptors (Table 1) and closely mirrored the GLP-1R and GIPR cAMP profiles of tirzepatide (Fig. 4). Notably, TC4’s GIPR potency was subtly enhanced compared to that of tirzepatide (p = 0.0186, Fig. S3). While balanced agonism in cell based bioassays is often considered predictive of in vivo efficacy,³⁵ TC4’s profile may confer advantages. The construct exhibits relatively uniform activity losses at GLP-1R, GcgR, and Y2R while maintaining robust GIPR agonism, which has been implicated in mitigating nausea associated with GLP-1R and Y2R activation, as mentioned earlier.^{15, 30} Furthermore, GIPR agonism has been shown to alleviate nausea in patients undergoing chemotherapy, reinforcing the potential therapeutic promise of TC4’s profile.¹⁶

Figure 4.

(a) β-arrestin-2 recruitment to GLP-1R measured using BRET. Experiments were performed using HEK 293T cells that were transiently transfected with GLP-1R-Rluc8 and the GFP²-β-arrestin-2 (R393E, R395E) fusion proteins. (b) AlphaFold 3 modelling showing ligand TC3 (Aib2Ala, Aib20Ala, and a free lysine at position 17) in complex with the GLP-1R, with TC3 residue Arg44 projecting away from the receptor. (c) A cartoon representation of the TC3 residue Arg44 in the ligand-receptor complex. Substituting R44 with NMeR disrupts βArr2 recruitment, reducing efficacy to 18% compared to GLP-1 (see panel (a) and Table 2). (d) Leucine 13 in TC3 may form important interactions when this ligand’s N-terminus binds to the GLP-1R pocket. Substitution of L13 with αMeL (together with NMe44, in chimera TC3f) impacts ligand interactions with the GLP-1R binding pocket. This may explain partially recovered β-arrestin-2 recruitment efficacy of TC3f compared to TC3c (panel (a) and Table 2). (e) Substitutions Lys16Arg and Aib20Ala in TC4 (vs. the TC3 chimera) markedly affect ligand interactions with the GLP-1R binding pocket. This may explain severely diminished β-arrestin-2 recruitment efficacy of TC4 at the GLP-1R (panel (a) and Table 2). (f) Cartoon representation of demonstrated tunable β-arrestin-2 signaling bias induced by tetra-peptide chimeras at the GLP-1R. Percentages shown in panels (c-e) represent relative efficacies in recruiting β-arrestin-2 to the GLP-1R (compared to efficacy of GLP-1 = 100%). Data shown in panel (a) represent the mean ± SEM of at least 3 independent experiments

Tetra-agonists induce biased signaling at the GLP-1R

The ability of ligands to stabilize distinct conformations of G protein-coupled receptors (GPCRs) can lead to selective engagement of intracellular signaling pathways, a phenomenon known as biased signaling.^{45, 46} Biased agonism at the GLP-1R is of particular interest, as ligands that preferentially stimulate cAMP production while reducing β-arrestin recruitment may sustain receptor signaling while minimizing receptor internalization and desensitization. Studies have demonstrated the advantages of cAMP-biased agonism as a means of enhancing glycemic control and weight loss.^{47, 48} Tirzepatide, for example, functions as a biased GLP-1R agonist, exhibiting cAMP signaling with reduced β-arrestin-2 (βArr2) recruitment, which may contribute to its prolonged activity despite reduced potency relative to native GLP-1.⁴⁹ Given the pharmacological implications of signaling bias, we sought to determine the βArr2 recruitment profiles of selected compounds among our tetra-agonist constructs to assess their potential for differential intracellular signaling.

To quantify ligand-induced βArr2 recruitment at the GLP-1R, we performed bioluminescence resonance energy transfer (BRET) assays using HEK293T cells transiently co-transfected with cDNAs encoding the fusion proteins GLP-1R-RLuc8 and GFP2-β-arrestin-2 (R393E, R395E).^{45, 48} Concentration-response curves were generated by incubating compounds with these transfected cells. Rluc8 substrate coelenterazine 400a was then added and incubated for 20−40 min before BRET signal (I₅₁₅ nm/I₃₉₅ nm) was measured and normalized to native GLP-1 activity (Table 2).

Table 2.

Potency and efficacy (% maximum response to GLP-1) of βArr2 recruitment at the GLP-1R.

β-arrestin-2
Peptidea	pEC50b	Max response (% GLP-1)c	n d
GLP-1	−7.42 ± 0.08	100	6
Semaglutide	−7.57 ± 0.05	92 ± 5	4
Tirzepatide	−7.57 ± 0.17	18 ± 2	6
Retatrutide	−7.62 ± 0.07	58 ± 5	6
GEP44	−7.67 ± 0.07	66 ± 4	4
TC0	−7.67 ± 0.09	90 ± 5	4
TC1	−7.55 ± 0.05	110 ± 7	3
TC2	−7.57 ± 0.21	43 ± 7	3
TC3	−7.29 ± 0.11	120 ± 5	4
TC3c	−7.14 ± 0.19	18 ± 2	6
TC3f	−7.51 ± 0.06	43 ± 2	4
TC4	−7.10 ± 0.18	14 ± 2	6

^aNomenclature of evaluated peptides corresponds to that used in Fig. 2 and Table 1.

^bLigand-induced β-arrestin-2 recruitment at the GLP-1R was measured by BRET assay. pEC₅₀ = −log(EC₅₀) determined in independent experiments (see SI). Values reported are the mean ± SEM.

^cMaximal response (efficacy) of each compound was normalized to the maximal response induced by GLP-1 (=100%) and the minimum response induced by vehicle (=0%).

^dNumber of independent experiments.

Consistent with prior reports, tirzepatide exhibited strong signaling bias, and recruited βArr2 with only 18% of the maximal efficacy of GLP-1.⁴⁹ Semaglutide showed minimal signaling bias vs. native GLP-1, while retatrutide displayed a modest reduction in maximal βArr2 recruitment (58% of GLP-1).¹⁸ We then examined the arrestin recruitment profiles of our chimeric constructs, beginning with TC0 and TC1, neither of which displayed significant signaling bias relative to GLP-1, while GEP44 showed a modest reduction in βArr2 related activity.⁵⁰

Among our tetra-agonists, TC2 exhibited a distinctly biased profile, with a significant reduction in βArr2 recruitment (43% of GLP-1) while maintaining potent cAMP signaling. Conversely, TC3 demonstrated a slight increase in βArr2 recruitment (120% of native GLP-1), suggesting that variations in lipidation or sequence composition may modulate receptor binding and downstream signaling. The influence of lipid type and position, as well as the presence of Aib at position 20 in TC3, may contribute to these differences in arrestin recruitment.

The structural basis for ligand-induced bias was further explored by introducing targeted substitutions within the TC3 scaffold. Remarkably, a single R44NMeR substitution in TC3c drastically reduced βArr2 recruitment (18% of GLP-1), despite the fact that residue 44 is distant from the N-terminal segment of TC3 which determines GLP-1R agonism. It is possible that the R44NMeR substitution disrupted a not yet defined interaction of this peptide with the extracellular receptor domain, thereby hindering some conformational GLP-1R change that facilitates β-arrestin mediated pathways (Table 2, Fig. 4a). To investigate this hypothesis, we employed AlphaFold 3 modeling,⁵¹ which confirmed that the “GGPS” spacer is predicted to effectively separate the Y2R-binding region from the GLP-1R interaction site. This finding supports the idea that residue 44 points away from the receptor, and that the R44NMeR modification may alter receptor-ligand binding dynamics via extracellular interactions that are distinct from the membrane embedded GLP-1R agonist pocket (Fig. 4b).

Further analysis revealed that introducing α-Me-L-leucine (αMeL) at position 13 in TC3f partially restored βArr2 recruitment compared to TC3c. This suggests that αMeL may stabilize a receptor conformation upon ligand binding that favors arrestin engagement (Fig. 4b, d). Importantly, TC3f retained cAMP potency comparable to TC3c while exhibiting significantly increased βArr2 recruitment (Fig. S3b). This observation highlights the potential of α-methyl-substituted residues as a tool for tuning biased signaling properties in GLP-1R agonists (Fig. 4f).

TC4, the final construct examined, displayed a signaling bias profile similar to tirzepatide and TC3c, with only 14% βArr2 recruitment relative to GLP-1 (Fig. 4a, e, Table 2). This result was particularly intriguing, as TC4 was generated via Lys16Arg and Aib20Ala substitutions within the TC3 scaffold. Notably, TC4 also exhibited enhanced agonism at GcgR and Y2R (Table 1), further distinguishing it from tirzepatide in its receptor selectivity. The striking resemblance between the signaling profiles of TC4 and tirzepatide reinforces the notion that specific residue modifications can be leveraged to fine-tune receptor engagement and intracellular response (Fig. 5a, b).

Figure 5.

Comparison of compound-stimulated receptor-mediated signaling via the cAMP and βArr2 pathways. (a) Normalized potencies of GLP-1, semaglutide, tirzepatide, retatrutide, and TC4 for GLP-1R mediated cAMP signaling. Data were obtained from Table 1 and Table S1, with GLP-1 potency defined as 100%. (b) Normalized maximal compound-induced βArr2 recruitment (efficacies) at the GLP-1R; data were normalized to the efficacy of GLP-1. (c) Correlation of GLP-1R mediated cAMP production potency of indicated compounds (fold-shift vs. potency of native GLP-1) vs. corresponding efficacies to induce GLP-1R mediated βArr2 recruitment (normalized to efficacy of native GLP-1 = 100%). Data were transformed from Table 1 and from Table 2. Compound TC2 is highlighted (*) as it is the only peptide that is acylated with a monoacid lipid instead of with a diacid lipid.

Our findings indicate that bias toward cAMP signaling versus βArr2 recruitment does not follow a strictly predictable pattern across different tetra-agonists. A comparative analysis of relative cAMP production and βArr2 recruitment suggests that increased cAMP potency does not always correlate with enhanced arrestin engagement. This is exemplified by the contrasting signaling profiles of TC2 and TC3, as well as the differential bias observed between TC3c and TC3f (Table 2, Fig. 5c, Fig. 7). The complexity of biased signaling at the GLP-1R and highlights the importance of structural determinants—such as lipidation, residue modifications, and ligand-receptor interactions—in shaping receptor bias. Taken together, these insights provide a framework for rationally engineering GLP-1R agonists with tailored signaling properties.

Figure 7.

Normalized potencies of agonists to stimulate (GLP-1R, GIPR, GcgR) or inhibit (Y2R) cAMP-dependent signaling via different receptors, illustrating distinct inter-receptor balance profiles. Data were derived from Table 1 and Table S1, and are expressed as a percentage of corresponding native peptide agonist potencies (GLP-1, GIP, Gcg, or PYY, each defined as 100%). No potencies at some receptors are indicated where no or marginal agonism was apparent (EC₅₀ > 1μM if detectable). Among tetra-chimeric (TC) compounds, relatively balanced agonist activities at all four target receptors were observed with TC2 whereas typically, potency preference for one or several of the targets occurred with lower activity at others (partial balance / imbalance). Data represent the mean ± SEM of at least 3 different experiments.

Tetra-agonists bind albumin and are stable to DPP4 action

To begin to further assess the biological implications of lipid acylation of the TC2 and TC3 constructs, we determined the half-maximal effective concentration (EC₅₀) at the glucagon-like peptide-1 receptor (GLP-1R) in the presence and absence of 2% human serum albumin (HSA) using a previously established cAMP-dependent luciferase reporter bioassay.^{8, 11, 35} Co-incubation with HSA resulted in a pronounced reduction in agonist potency. This phenomenon suggests that lipidation facilitates the formation of an HSA-peptide complex thereby hindering receptor-ligand interactions and signaling, and that the degree of HSA-induced loss of compound potency is indicative of albumin binding affinity (Table S2). Among the constructs evaluated, semaglutide exhibited the most substantial HSA-induced potency shift (1,300-fold), while TC2 and TC3 displayed comparatively moderate changes (130- and 140-fold, respectively). This trend aligns with prior findings, wherein the linker regions of lipidated diacids were shown to influence receptor binding ratios at GLP-1R in the presence and absence of HSA. Notably, differential binding affinities were reported for C20 and C18 diacids at equivalent linker positions, yet ultracentrifugation analyses indicated similar HSA interactions and comparable circulation half-lives in rodent models.⁸ These observations suggest that although the GLP-1R EC₅₀ shift with vs. without HSA provides insight into albumin binding, it may not fully predict pharmacokinetic behavior. Given the clear EC₅₀ shift observed for TC3 (± 2% HSA), it is likely that this analogue, along with other diacid-containing constructs developed in this study, would achieve extended circulation times in the blood stream, similar to what has been observed with other compounds that are currently in the clinic or in late-stage clinical trials.

Having established that the lipid side chain confers robust albumin binding, we next examined the constructs’ stability against degradation by dipeptidyl peptidase-4 (DPP4). Prior studies have emphasized that, in addition to lipidation, DPP4-resistant amino acid sequence elements are critical for extending peptide half-life in vivo.⁸ To evaluate the susceptibility of our analogues to DPP4-catalyzed cleavage, we employed a cAMP-based luciferase assay, leveraging the observation that the DPP4-truncated GLP-1(9–36 amide) exhibits a >1,000-fold loss in receptor potency. Accordingly, incubation of GLP-1 and our constructs with either vehicle or DPP4 enabled the quantification of potency shifts that result from enzyme catalyzed hydrolysis and inactivation. As expected, DPP4 treatment significantly diminished the potency of GLP-1 (Fig. 6a, Table S3), consistent with prior reports demonstrating its rapid conversion to GLP-1(9–36 amide) within minutes under assay conditions.⁶ In contrast, TC2, TC3, and TC4 exhibited no significant change in activity upon DPP4 exposure, confirming their resistance to proteolysis (Fig. 6b–d, Table S3). This result aligns with previous findings that substitution of the second residue in the DPP4 cleavage motif (P1 position to the scissile bond) with aminoisobutyric acid (Aib) confers complete enzymatic protection in substrates such as GLP-1 and GIP.

Figure 6.

Stability comparison of compounds towards DPP4-catalyzed peptide hydrolysis and inactivation. Native GLP-1 is susceptible to enzyme-induced inactivation as reflected by a 430-fold potency loss, while all TC constructs are refractory to DPP4 action. (a) GLP-1 (n = 3), (b) TC2 (n = 3), (c) TC3 (n = 3), and (d) TC4 (n = 1) were preincubated with DPP4 (open symbols) or with vehicle (closed symbols) for 18 hours at 37 °C. Subsequently, cAMP dependent signaling was determined using HEK-293 C34L cells stably transfected with GLP-1R and CRE_6x-luciferase cDNAs. Data represent mean ± SEM.

DISCUSSION

The current study demonstrates feasible strategies toward engineering unprecedented monomolecular tetra-chimeric peptides, compounds that were designed to potentially enhance therapeutic reduction of excessive body weight. For the first time, we generated multi-agonists that integrate anorectic PYY functionality with the established synergistic triagonism of GLP-1, GIP, and glucagon that is currently emerging in late-stage phase 3 clinical studies as a very effective treatment of obesity.¹⁹ While the therapeutic potential of our next-generation tetra-agonists remains to be investigated in future animal and clinical studies, underlying considerations that lead to the rational design of such molecules support the promise of further exploration.

Tetra-agonists as described here are designed to combine four different hormonal functions that mostly synergize to induce weight loss but in part also counter-act each other to attenuate unwanted effects. The main foundation on which our tetra-agonists are built – similar to most existing or emerging dual or tri-agonist drugs for treating diabetes and obesity – is GLP-1, given its stand-alone glucoregulatory and satiety-inducing functions. In more recent multi-agonists, the function of GLP-1 is often combined with that of GIP (enhancement of GLP-1 induced functions via distinct receptors, synergistic modulation of fatty acid metabolism) and/or that of glucagon (enhanced energy expenditure to boost weight loss, while concomitant liability to induce hyperglycemia is counteracted by GLP-1+GIP). In tetra-agonists, we now add the function of PYY to induce Y2R activation.

Extensive studies of PYY/Y2R physiology predict that concomitant activation of this mechanism will further support weight loss by a combination of GLP-1 and GIP-like complementary functions that are however triggered via a distinct set of distinct receptors, target cells, and mechanisms (Fig. 1). Supporting this idea, GLP-1R and Y2R dual-agonists have been shown to enhance appetite suppression beyond that of GLP-1R agonists alone,^{27, 29} while co-administration of GIPR and Y2R agonists has been reported to mitigate nausea, a common adverse effect associated with both GLP-1R and Y2R mono-agonists.³⁰ In contrast to glucagon, PYY/Y2R activation does not interfere with blood glucose homeostasis as a potential liability when included in a multi-agonist. Of note, both PYY and GLP-1 tend to trigger nausea at higher concentrations, albeit through different neuronal circuits. However, nausea induced by either of these hormones as a frequently encountered side effect is effectively counteracted when these hormones are combined with GIP.^{15, 30}

In view of this complex interplay between synergistic and mutually attenuating functions of proposed multi-functional agonists, it is important to optimally tune the balance among included hormonal functions. Feasibility to do this when engineering tetra-chimeric (TC) peptides is illustrated in Fig. 7, however selecting an optimal balance to induce weight loss and enable tolerated dose escalation while avoiding side effects, awaits systematic compound comparison in future in vivo studies. Comparison of TC compounds in Fig. 7 with Tirzepatide (clinically effective GLP-1/GIP dual agonist) and retatrutide (promising GLP-1/GIP/glucagon tri-agonist in late phase 3 studies) suggests that maximized GIPR potency (associated with no known side effect, but linked to attenuating nausea triggered by GLP-1 and PYY) may be a desirable feature when prioritizing compounds for in vivo evaluation of potential therapeutic promise. One may also predict that inclusion of PYY function in a tetra-chimeric agonist will enable reduced reliance on GLP-1 and glucagon activity for inducing weight loss, which could diminish the likelihood of associated dose-limiting side effects (nausea and hyperglycemia, respectively).

From a peptide engineering perspective, we found that inter-receptor potency balancing of TCs can be achieved by peptide lipidations, by C-terminal modifications, as well as by minor modifications at the N-terminus. Our work highlights how modest sequence variations near the N-terminus impact signaling dynamics, consistent with previous findings that this region is a key determinant of biased agonism across GLP-1R, GIPR, and GcgR.^{2, 3, 6} We have previously shown that strategic alkylation of the N-terminal amine can differentially modulate cAMP production at distinct target GPCRs, either enhancing or attenuating receptor activation, depending on the chemical modification employed. Extending this principle to multi-agonists offers a powerful approach for shifting receptor balance, particularly in the context of designing ligands with improved tolerability and metabolic efficacy. This highlights the structural flexibility of our platform and supports the feasibility of further pharmacological tuning beyond the range shown in Fig. 7.

In fact, we found that tunability of TCs in this way is not limited to adjusting inter-receptor potency balance, but also extends to adjusting pathway-dependent biased signaling (cAMP- vs. β-arrestin-mediated) at the critical GLP-1R (Fig. S3c). Multiple reports now point to a requirement of cAMP-biased GLP-1R signaling for maximizing weight loss efficiency of cognate agonists and possibly for minimizing induction of nausea as a side effect, e.g. as observed with the clinically used dual agonist Tirzepatide. Several of our new TC constructs show similar cAMP-biased signaling (with little agonist-induced induction of β-arrestin recruitment) as observed with Tirzepatide (Figs. 4 and 5), thereby supporting the translational promise of such compounds.

A key design element of our TC constructs is the insertion of a “GGPS” spacer derived from the tail segment of a GLP-1 related natural peptide, exendin-4. This spacer effectively separates the N-terminal domain of TCs corresponding to GLP-1/GIP/Gcg agonist sequence from the C-terminal TC domain that determines PYY agonism at Y2R. We believe that such separation contributes to our ability to independently tune agonism at each of the corresponding target receptors, to be further optimized with guidance from compound-induced weight loss efficiency to be determined in future in vivo experiments. Balancing of TCs may be achieved without necessarily requiring substitution of noncanonical amino acids that are commonly incorporated into multi-agonists.^{18, 52, 53}

In anticipation of future exploration of TC pharmacokinetics and therapeutically induced weight loss in vivo, we conducted early mechanistic studies of key compounds to estimate the likelihood that extended half-lives in the blood stream can be achieved. We noted that lipidated variants of these compounds showed evidence of high affinity binding to albumin, a well-established mechanism of peptide drug protraction via delaying renal elimination and protection from enzymatic degradation that enabled most GLP-1 related medications currently in the clinic, including semaglutide and tirzepatide. Furthermore, we observed that several of our compounds were protected from rapid enzymatic degradation by the ubiquitous frontline protease, which rapidly inactivates endogenous GLP-1 by N-terminal cleavage. Such protection of our TCs is likely attributable to a combination of N-terminal peptide modification in combination with lipidation (as above). Based on this preliminary evidence, it appears that key TCs from this study are ready for evaluation in animal experiments and/or are ready for further pharmacokinetic optimization if needed.

Finally, we considered feasibility of efficient future large-scale production in case one or several of our TCs may emerge as a viable lead for further development. Application of this criterion has come to the forefront with the success of GLP-1 related peptide drugs semaglutide and tirzepatide in the clinic where ensuring sufficient supply to match demand and contain production cost have become a limiting factor. While, with Semaglutide, semi-recombinant production is feasible,⁴⁵ Tirzepatide (a 39-amino-acid molecule) needs to be generated by an involved hybrid solid-phase and liquid-phase chemical synthesis approach due to the presence of Aib at position 13 of this peptide and multiple lysines that preclude orthogonal post-synthetic lipidation.³³ The latter approach imposes substantial constraints on manufacturing, placing pressure on Fmoc-amino acid supply chains, requiring the use of non-green solvents, and increasing production costs. Furthermore, even higher order multi-agonists such as one that was recently patented to co-activate the GLP-1R, GIPR, amylin and calcitonin receptors and includes more the 60 amino acids would be even more difficult to chemically synthesize at large scale in case drug development is successful.⁵³ Given the rapid expansion of the GLP-1R agonist field with increasingly complex pharmacology, recombinant synthesis must be prioritized as a scalable and sustainable alternative. With these challenges in mind, we limited our TCs to 45 residues and specifically designed TC4 as a scaffold amendable to recombinant expression. In this construct, a recombinant template could be expressed up to position 3 with an additional C-terminal glycine (position 46), after which C-terminal amidation could be introduced via established chemistries,⁴⁶ lipidation could be site-specifically incorporated at Lys17, and an N-terminal dipeptide (Tyr-Aib) could be coupled. This approach parallels the semi-recominant strategy employed for semaglutide,⁴⁵ offering a pathway toward scalable, cost-effective production of tetra-agonists.

Taken together, our findings establish a new paradigm for designing unimolecular, tetra-chimeric agonists with expanded and tunable pharmacology, and one that can be adapted for feasible large-scale production. Such compounds, by co-leveraging complementary mechanisms in addition to those activated by GLP-1 and other glucagon-related peptides, have the potential to base future weight loss therapies on a broadened range of multi-pronged mechanisms. This, in turn, may provide a viable alternative therapeutic option for the significant number of patients that do not respond to currently considered GLP-1 based mono-, dual, or tri-agonists, or abandon treatment with these compounds due to intolerable side effects including nausea and vomiting. By integrating rational design principles with innovative synthetic strategies, our work provides a roadmap for the next generation of peptide-based therapeutics with possible applications in metabolic disease and beyond.

Supplementary Material

Experimental procedures, peptide synthesis and purification, peptide potency with SEM for all data reported in Table 1 (Table S1), HSA incubation assays (Table S2), DPP4 incubation assays (Table S3), the luciferase reporter assays, and β-arrestin-2 recruitment assays. HPLC and mass spectrometry tables confirming purity and identity of peptide variants (Table S4).

ABBREVIATIONS USED

GLP-1

Glucagon-like peptide-1

GIP

Glucose-dependent insulinotropic polypeptide

Gcg

glucagon

PYY

Peptide YY

Y2R

neuropeptide Y2 receptor

DPP4

dipeptidyl peptidase-4

Aib, X

2-aminoisobutyric acid

αMeL, L^α

α-methyl-L-Leucine

T2D

type 2 diabetes

TC

tetra-chimeric

DA

diacid

NMeR

N-methyl arginine

ɣGlu, ɣE

ɣ-L-glutamic acid

OEG, O

8-amino-3,6-dioxaoctanoic acid

βAla, βA

β-alanine

Glossary

Lipid side chain abbreviations are lysine side chain acylated, such as

lipid 1

OEG-OEG-ɣGlu-C18DA

lipid 2

OEG-OEG-ɣGlu-C20DA

lipid 3

ɣGlu-C16 acyl

lipid 4

OEG-ɣGlu-C18DA

O_2X-βA-ɣE-C18DA

OEG-OEG-βala-ɣGlu-C18DA

O-βA-ɣE-C18DA

OEG-βala-ɣGlu-C18DA

lipid 5

βala-OEG-ɣGlu-C18DA